Cross-age face recognition method based on feature fusion

Abstract

The invention discloses a cross-age face recognition method based on feature fusion, and the method comprises the steps: extracting eye shallow stable features of a face, extracting age-independent deep identity features through an orthogonal decomposition and joint learning mode, carrying out the fusion and dimension reduction of the two types of features in a full connection layer, and achieving the classification based on an Arcface loss function. According to the method, the LBP features of the eye region of the human face are added into the features for recognition, self-quotient enhancement is carried out on the eye region, and discrete wavelet transform noise reduction is carried out on the enhanced picture. Therefore, the obtained features are kept stable along with the change of age, and the robustness of LBP local binary mode coding on factors such as illumination in the picture is also high.

Description

technical field

[0001] The invention relates to the technical field of computer image processing, in particular to a feature fusion-based cross-age face recognition method. Background technique

[0002] Cross-age face recognition has a wide range of applications, including looking for lost children and escaped criminals. Its most important feature is the age difference. It requires the model to judge whether the given two pictures are of different ages when the pictures of the same person are of different ages. same person. The difficulty of cross-age face recognition lies in the large intra-class differences. Compared with other face attributes, age changes are more complicated and have a huge impact on faces. It is easy to have intra-class differences greater than inter-class differences. . Considering the difference between cross-age face recognition and general face recognition and the wide application of cross-age face recognition, it is necessary to study cross-age f...

Images